New optimality conditions for bilevel programs by using an exact separation principle (Q2192985)

The article is aimed at deriving new optimality conditions for bilevel optimization problems with inequality constraints where a part of the variables belongs to the solution set of a parametric optimization problem formulated for the other variables. It is assumed that the objective and the constraint functions are locally Lipschitz continuous. The authors use the exact separation theorem proved by \textit{X. Y. Zheng} et al. [Optimization 66, No. 7, 1065--1077 (2017; Zbl 1407.90344)] to obtain exact optimality conditions for the bilevel optimization problem under consideration in terms of normal cones to the feasible set, namely the Fréchet normal cone and the Mordukhovich (limiting) normal cones. The new optimality conditions are proved in the form of two theorems, one of which gives a sequential optimality condition without any qualification, and the other -- under a nonsmooth sequentially Abadie constraint qualification. Both conditions are derived in terms of Fréchet subdifferentials of the initial data.